An Efficient Solution to the Mutual Exclusion Problem Using Weak Semaphores

We present a simple solution to the mutual exclusion problem with no lockout that uses weak binary semaphores. Previously, Morris [4] presented a solution to this problem that uses three binary semaphores and two variable-size shared variables. A simpler solution was given in [5], where just two semaphores and two shared singlebit variables are used. (In some earlier versions, an incorrect one-bit solution was given.) The algorithm given in this paper also uses two semaphores and two shared single-bit variables, but it solves an additional open problem by being correct even if the number of processes is unbounded but finite. (Morris uses the term 'finite' when 'bounde~l' is dearly meant.) The algorithm makes use of a form of semaphores we call 'weak'. The semantics of these semaphores is that if a process performs a V operation and some other process is waiting on that semaphore, then the first process is not allowed to immediately perform another P operation on that semaphore. Instead, one of the wait-